蚁群优化算法
MOSFET
材料科学
响应面法
结温
碳化硅
有限元法
人工神经网络
电子工程
晶体管
机械工程
工程类
热的
结构工程
计算机科学
复合材料
电气工程
算法
物理
电压
机器学习
气象学
作者
Yichen Qian,Fengze Hou,Jiajie Fan,Quanya Lv,Xuejun Fan,Guoqi Zhang
标识
DOI:10.1109/ted.2021.3077209
摘要
A new panel-level silicon carbide (SiC) metal oxide semiconductor field effect transistor (MOSFET) power module was developed by using the fan-out and embedded chip technologies. To achieve the more effective thermal management and higher reliability under thermal cycling, a new optimization method called Ant colony optimization-back propagation neural network (ACO-BPNN) was developed for optimizing SiC modules, and contrast it with the Response Surface Method (RSM). First, the heat dissipations of SiC MOSFET with different redistribution layer (RDL) materials were simulated through the ANSYS finite element simulation. Then, the RSM was adopted to design the experiments for optimization. Third, the optimized design considering both junction temperature and thermal-mechanical stress is obtained using RSM and ACO-BPNN. The results show that: 1) compared with nano-silver, copper has a relatively good heat dissipation effect, but nano-silver has a better thermodynamic performance, and 2) ACO-BPNN can provide more accurate optimization results without having to construct a fitness function like RSM. After optimization, the thermal management and the thermal-mechanical stress can be improved by about 3% and 11%, respectively.
科研通智能强力驱动
Strongly Powered by AbleSci AI